Designing Data-Intensive Applications

About halfway through. I'm particularly interested in the sections on replication, partitioning, and consistency.

My notes One of the clearest explanations I've found of distributed-system tradeoffs. Rather than prescribing one architecture, it gives names to the choices that show up again and again.

Ideas I'm thinking about Reliability requires thinking about what can fail. Replication introduces questions about consistency. A good design starts by making its assumptions explicit.